[hip-kernel-provider] Implement RMSNorm backward kernels and RMSNorm channel last support#7702
Merged
saikubairkota merged 13 commits intoMay 29, 2026
Conversation
sbalint98
reviewed
May 26, 2026
Contributor
sbalint98
left a comment
sbalint98
left a comment
There was a problem hiding this comment.
Looks good to me with one minor improvement suggestion
saikubairkota
force-pushed
the
streamhpc/saikubairkota/rmsnorm-fwd-channel-last-en-bwd-kernel
branch
from
7002769 to
6d259d3
Compare
…lementation of channel last support for RMSNorm forward kernel
… a common header file
…, stride, & normalize dim calculation headers to common header for reuse
…plicability checks where affine tensors have to accept all types instead of stat tensors
…el dim when calculating outer size for channel-last layouts
…r RMSNorm backward kernel
…ect changes in kernel compilation & execution
saikubairkota
force-pushed
the
streamhpc/saikubairkota/rmsnorm-fwd-channel-last-en-bwd-kernel
branch
from
6d259d3 to
e704121
Compare
EwanC
approved these changes
May 27, 2026
Contributor
EwanC
left a comment
EwanC
left a comment
There was a problem hiding this comment.
LGTM, other than a test naming convention issue that CI picked up
TestRMSnormBwdPlan.CompileBFp16SetsCorrectDefines [FAIL]
-> Test case name should not contain keywords: Fp16. These belong in the test suite name.
TestRMSnormBwdPlan.CompileFp16SetsCorrectDefines [FAIL]
-> Test case name should not contain keywords: Fp16. These belong in the test suite name.
TestRMSnormBwdPlan.CompileFp32SetsCorrectDefines [FAIL]
-> Test case name should not contain keywords: Fp32. These belong in the test suite name.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
3 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Caution: This PR should be merged only after this PR is merged.
Motivation
This PR implements the RMSNorm backward kernels and RMSNorm channel-last support for both forward and backward operations in the hip kernel provider.
Technical Details
RMSnormBwdPlanto compile and launch the kernels.RMSnormFwdandRMSnormBwdoperations.Test Plan
Build the plugin and run the unit and integration tests with
ninja check.Test Result
All unit and integration tests pass successfully on an MI210.
Submission Checklist